Mothering alone: cross-national comparisons of later

Research report
Mothering alone: cross-national comparisons of
later-life disability and health among women who
were single mothers
Lisa F Berkman,1,2 Yuhui Zheng,3 M Maria Glymour,4,5 Mauricio Avendano,4,6
Axel Börsch-Supan,6 Erika L Sabbath1,7
▸ Additional material is
published online only. To view
please visit the journal online
(http://dx.doi.org/10.1136/jech2014-205149).
1
Harvard Center for Population
and Development Studies,
Cambridge, Massachusetts,
USA
2
Departments of Social and
Behavioral Sciences;
Epidemiology; and Global
Health and Population,
Harvard T.H. Chan School of
Public Health, Boston,
Massachusetts, USA
3
School of Public Policy and
Management, Tsinghua
University, Beijing, China
4
Department of Social and
Behavioral Sciences, Harvard
School of Public Health,
Boston, Massachusetts, USA
5
Department of Epidemiology
and Biostatistics, University of
California, San Francisco,
California, USA
6
London School of Economics
and Political Science, LSE
Health, London, UK
7
Munich Center for the
Economics of Aging at the
Max Planck Institute for Social
Law and Social Policy, Munich,
Germany
8
School of Social Work, Boston
College, Chestnut Hill,
Massachusetts, USA
Correspondence to
Dr Lisa F Berkman, Harvard
Center for Population and
Development Studies, 9 Bow
Street, Cambridge, MA 02138,
USA;
lberkman@hsph.harvard.edu
Received 23 October 2014
Revised 24 February 2015
Accepted 28 February 2015
To cite: Berkman LF,
Zheng Y, Glymour MM,
et al. J Epidemiol
Community Health Published
Online First: [ please include
Day Month Year]
doi:10.1136/jech-2014205149
ABSTRACT
Background Single motherhood is associated with
poorer health, but whether this association varies
between countries is not known. We examine
associations between single motherhood and poor laterlife health in the USA, England and 13 European
countries.
Methods Data came from 25 125 women aged 50+
who participated in the US Health and Retirement Study,
the English Longitudinal Study of Ageing and Survey of
Health, Ageing and Retirement in Europe. We tested
whether single motherhood at ages 16–49 was
associated with increased risk of limitations with activities
of daily living (ADL), instrumental ADL and fair/poor selfrated health in later life.
Results 33% of American mothers had experienced
single motherhood before age 50, versus 22% in
England, 38% in Scandinavia, 22% in Western Europe
and 10% in Southern Europe. Single mothers had higher
risk of poorer health and disability in later life than
married mothers, but associations varied between
countries. For example, risk ratios for ADL limitations were
1.51 (95% CI 1.29 to 1.98) in England, 1.50 (1.10 to
2.05) in Scandinavia and 1.27 (1.17 to 1.40) in the USA,
versus 1.09 (0.80 to 1.47) in Western Europe, 1.13 (0.80
to 1.60) in Southern Europe and 0.93 (0.66 to 1.31) in
Eastern Europe. Women who were single mothers before
age 20, for 8+ years, or resulting from divorce or nonmarital childbearing, were at particular risk.
Conclusions Single motherhood during early-adulthood
or mid-adulthood is associated with poorer health in later
life. Risks were greatest in England, the USA and
Scandinavia. Selection and causation mechanisms might
both explain between-country variation.
INTRODUCTION
Single motherhood—the experience of parenting
without a marital partner—is associated with
increased risk of health problems, including poor
self-rated health (SRH), adverse cardiovascular risk,
poor mental health and increased mortality.1–11
Prior studies have focused primarily on contemporaneous associations between single motherhood and
health, but few studies have examined the ‘long arm
of single motherhood’, or how single motherhood
during early and mid-adulthood relates to health
and functioning at older ages. Recent birth cohorts
are increasingly likely to have experienced a spell of
single parenting,12 but little research links single
parenting to health in later life.
Except for several comparative studies of two or
three countries,4 6 9 no studies have systematically
examined whether associations between single
motherhood and health vary across countries; this
question is important for several reasons. Single
motherhood is associated with poverty in most
societies, but more so in the USA than in
Europe.13 14 This may lead to different mechanisms
of selection into lone motherhood between countries. Particularly in Southern European countries,
strong social and family networks may offset some
negative effects of single motherhood. Single
mothers’ risk of poverty, for example, may be offset
by family support. Family policies aiming to encourage women to combine motherhood with labour
force participation in the UK and the European continent may have positive effects, but they may also
have unintended consequences. For example, feminist welfare state theories suggest that family policies
may in fact reinforce women’s roles as unpaid caregivers or encourage part time paid work.15–17
We hypothesise that women experiencing an
episode of single motherhood before age 50 have
worse health at older ages than married mothers,
and that single motherhood is most damaging in
countries with relatively weak social safety nets,
such as the USA and England. Building on a lifecourse model of health, we assess cumulative
effects of single parenting as a risk factor for
poorer functioning and health at older ages. We
test these hypotheses using harmonised data from
population-based studies of older adults in the
USA, England and 13 continental European
countries.
METHODS
Data
We used three harmonised longitudinal surveys on
health and ageing: the US Health and Retirement
Study (HRS), English Longitudinal Study of Aging
(ELSA), and Survey of Health, Aging, and Retirement
in Europe (SHARE), which represents 21 continental
European countries (13 of which collected life
history data essential for these analyses). These
surveys are described in detail elsewhere.18–21 Briefly,
each study conducts biennial assessments of nationally representative samples of non-institutionalised
adults age 50+. HRS was implemented in 1992,
ELSA in 2002 and SHARE in 2004. Survey comparability is discussed elsewhere.22 We categorised the
13 European SHARE countries into four geographic
regions: Scandinavia (Denmark, Sweden), Western
Europe (Austria, France, Germany, Switzerland,
Belgium, the Netherlands), Southern Europe (Italy,
Spain, Greece) and Eastern Europe (Poland, the
Berkman LF, et al. J Epidemiol Community Health 2015;0:1–8. doi:10.1136/jech-2014-205149
1
Research report
Czech Republic). This categorisation is based on geographic and
cultural proximity and types of welfare regimes.
HRS data from 2004 and 2006 are included. ELSA data are
from 2004 and 2006; ELSA life history interviews regarding
childbearing and marriage were conducted in 2006, so we
include only ELSA respondents who completed 2006 interviews. SHARE includes respondents participating in the 2008
wave (SHARELIFE), which collected life histories on childbearing and marriage. SHARE participants were interviewed at least
once in 2004 or 2006. Response rates for the 2004 and 2006
HRS ranged from 75.3% to 91.4%.23 In ELSA, response rates
were 81% in 2004 and 69% for completing both main and life
history interviews in 2006.19 Overall SHARE rates were 52.5%
in 2004, 46% in 2006 and 61% in 2008, with country variation.20 Each survey provides individual-level sample weights,
used in descriptive and regression analyses, accounting for
sample design and non-response; weighted samples are nationally representative of target populations in each country by
survey year. We excluded women with no children by age 50.
The final analytic sample was 25 125 women aged 50+, with
42 830 observations total (17 866 from HRS, 6294 from ELSA
and 18 670 from SHARE). This study was approved by relevant
human subjects committees.
Outcomes
We examined three outcomes: limitations in activities of daily
living (ADLs), limitations in instrumental ADLs (IADLs) and
fair/poor SRH.24 25 ADL questions asked about bathing, dressing, eating, getting in/out of bed and walking across the room.
Participants were asked if they had any difficulty because of
physical, mental, emotional or memory problems. Response
options were binary (yes/no) in ELSA and SHARE. In HRS
there were two additional options: ‘don’t do’ or ‘can’t do.’
Individuals are classified as having any ADL limitation if they
reported ‘yes’ or ‘can’t do’. IADL questions asked about any difficulty with: making meals, shopping, making phone calls, medications and managing money. Those reporting ‘yes’ or ‘can’t do’
for any activity are classified as having an IADL limitation. SRH
is assessed by asking ‘Would you say your health is …’ with a
Likert scale response (excellent/very good/good/fair/poor). We
dichotomise SRH into fair/poor versus other.
Predictors
The key predictor of interest is single motherhood experience
between ages 15–49. A woman was considered a single mother
in any year when she had children under age 18 and was not
married. Each woman was asked to report all children’s birth or
adoption dates. Women were asked about beginning and ending
dates of each marriage. For each year between ages 15 and 49,
we created indicators for whether the participant had at least
one biological or adopted child under age 18 (based on child
birth year data) and for whether she was married (based on
whether that year fell between beginning and ending years of
any marriage reported). Information on non-marital partners
was not consistently collected, so is not included in main analyses, but is used in sensitivity analyses. Child and marriage variables were used to generate a binary indicator for whether a
woman ever experienced single motherhood before age 50. We
developed categorical indicators for duration of single motherhood (1–3; 4–7; 8–13; 14+ years), corresponding to quartiles
of single motherhood duration among those with any single
motherhood history. We further characterised types of single
motherhood (attributable to non-marital childbearing,
2
widowhood or divorce) and earliest age of single motherhood
(before age 20, 20–29, 30–39, 40+).
Statistical analysis
For each outcome, we estimate adjusted relative risks (RRs) and
95% CIs associated with single motherhood in each region
using modified Poisson regression models, which assume
Poisson distributions and use robust variances to correct for
error term misspecification. Modified Poisson regression permits
estimation of RRs with common binary outcomes26 if logbinomial models fail to converge, as occurred here.27–33
Because we had up to two observations per individual, we corrected SEs by clustering at the individual level to account for
correlations between repeated outcomes in the same woman.
Sensitivity analyses used one observation per woman.
For primary analyses, key independent variables were interactions between the six country/region dummy variables (USA,
England, Scandinavia, Western Europe, Southern Europe, Eastern
Europe) and an indicator of any single motherhood experience.
We additionally adjusted for covariates: assessment year, age, age
squared, educational attainment (secondary, primary or less, tertiary (reference)), number of children (one (reference), two, three
or more) and current marital status (married (reference) or not).
We allow effects of covariates to vary by country/region by including region-covariate interaction terms. We include country-level
fixed effects. We conduct Wald tests to assess whether RRs associated with single motherhood were equivalent across different
country/regions.34 We use α criteria of 0.05 and 0.10, for statistical
and marginal significance, respectively.
Next, we examine whether adjustment for current relative
income and wealth attenuates associations between single
motherhood and outcomes by region. For these models, we add
interactions of six country/region dummies with per-capita
household income and wealth quintiles. These metrics were generated by dividing income or wealth by square root of household size.35 We used country-specific and time-specific income
and wealth quintiles.
We investigate variation in observed associations by single
motherhood duration, type and age. Sample size limitations precluded interactions between those characteristics and the six
regional indicators. Models adjust for core covariates as in
primary analyses. Models for type and age of single motherhood were each adjusted for duration. Sampling weights were
used and robust variances clustered within individuals were estimated. While our preference was for region-specific analysis,
small sample sizes by regions meant we had limited power to
explore regional effects of duration and pathways into single
motherhood. We therefore focus on pooled analyses.
Analyses were conducted in Stata Special Edition, V.11
(StataCorp, College Station, Texas, USA).
RESULTS
History of single motherhood among women age 50+
In the USA, 32.8% of mothers aged 50+ had any single
motherhood experience between the ages of 15–49, compared
to 22.0% in England, 38.2% in Scandinavia and 10.2% in
Southern Europe (table 1). Divorce was the most common
reason for single motherhood. In European countries and
England, excluding unmarried women with partners from the
single motherhood definition, lifetime prevalence of single
motherhood decreased by less than four percentage points,
except in Scandinavia where it decreased by 11 percentage
points. In pooled analyses, we do not take partnership into
account since it is not available for all countries. In sensitivity
Berkman LF, et al. J Epidemiol Community Health 2015;0:1–8. doi:10.1136/jech-2014-205149
Research report
Table 1 Single motherhood experience among mothers aged 50+, by region
Ever single mother, No (%)
Single motherhood due to
Non-marital childbearing*
Widowhood
Divorce
Single motherhood at age
<20
20–29
30–39
40–49
Ever single mother without partner†, No (%)
USA
(n=17 866)
England
(n=6294)
Scandinavia
(n=2972)
Western Europe
(n=8576)
Southern Europe
(n=5305)
Eastern Europe
(n=1817)
5429 (32.8)
1337 (22.0)
1068 (38.2)
1687 (22.8)
472 (10.2)
312 (20.4)
1710 (8.8)
829 (4.8)
3494 (22.6)
353 (6.2)
183 (3.0)
875 (14.3)
527 (18.2)
74 (2.9)
559 (20.5)
606 (8.7)
300 (4.5)
870 (11.3)
129 (3.1)
193 (4.5)
157 (2.8)
94 (5.8)
76 (6.0)
154 (9.7)
868 (4.2)
2046 (12.5)
1558 (10.2)
957 (5.8)
NA
88 (1.6)
394 (6.8)
517 (8.5)
338 (5.2)
1106 (18.3)
32
121
148
171
405
26
113
103
70
275
122
467
314
165
714
(4.3)
(15.9)
(11.6)
(6.4)
(26.9)
154 (2.3)
612 (8.7)
553 (7.3)
368 (4.5)
1420 (19.3)
(0.9)
(2.8)
(2.9)
(3.7)
(8.8)
(1.0)
(8.3)
(6.8)
(4.2)
(17.8)
Mothers include women who have ever had any biological or adopted children. A woman was defined as having been “ever single mother” if, in any year when she was age 15–49,
she had a child when under the age of 18 years but was not married.
Number of observations and the percentages (in parentheses) are shown. Sampling weights were used for estimating percentages.
Countries in each regions are: Scandinavia (Denmark and Sweden), Western Europe (Austria, France, Germany, Switzerland, Belgium and the Netherlands), Southern Europe (Italy, Spain
and Greece) and Eastern Europe (Poland and the Czech Republic).
Data sources: Health and Retirement Study (HRS) in year 2004 and 2006; English Longitudinal Study of Ageing (ELSA) in year 2004 and 2006; Survey of Health, Ageing, and Retirement
in Europe (SHARE) in year 2004 and 2006 and SHARELIFE in year 2008.
*Mothers in the ‘non-marital childbearing’ category include the ‘never married’ group as well as women who reported one or more marriages, but were unmarried at the time of the
child’s birth.
†HRS does not collect information on partnership history.
analyses by region, however, we explicitly test associations of
partnership versus marriage with health.
Sample characteristics by single motherhood status
In every region, women with past experiences of single motherhood were younger, had lower income and wealth, and were
less likely to be married as older adults compared with consistently married mothers (table 2). In the USA and England, single
mothers were more likely to have primary education or lower.
Single motherhood was not associated with education in other
regions.
Are associations between single motherhood and
functioning and health similar across countries?
Single motherhood was associated with higher risk of ADL/
IADL limitations and fair/poor SRH in the USA and in England;
with ADL limitations and SRH in Scandinavia; but only with
SRH in Western Europe (table 3, Model I). For ADL limitations,
RRs associated with single motherhood were highest in England
(RR 1.51; 95% CI 1.29 to 1.77), followed by Scandinavia
(1.50; 1.10 to 2.05) and the USA (1.27; 1.14 to 1.40). RRs in
Western Europe, Southern Europe and Eastern Europe were
close to one and not significant. Wald tests provided marginally
significant evidence ( p=0.074) that coefficients for single
motherhood differed by region.
Single motherhood experience was associated with IADL limitations in England (1.66; 1.36 to 2.02) and the USA (1.27;
1.14 to 1.42) only. Wald tests showed RRs in England was significantly higher than RRs anywhere else. Single motherhood
was associated with higher risk of poor SRH in all regions
except Southern and Eastern Europe; associations were largest
in England (1.61; 1.43 to 1.81).
Do differences in income and wealth explain associations
between single motherhood and health?
Adjusting for income and wealth quintiles, RRs for any ADL
and IADL limitations were attenuated by more than 66% from
Model I and were no longer statistically significant in the US
(table 3, Model II). The RR for SRH in US single mothers was
substantially attenuated (from 1.32 to 1.16) but remained statistically significant after adjustment. In England, adjustment for
income and wealth modestly attenuated associations between
single motherhood and outcomes; all RRs remained statistically
significant. In Scandinavia, adjustment for income and wealth
modestly attenuated RRs for ADLs (1.50 to 1.40) and SRH
(1.20 to 1.12).
Duration, type and age of single motherhood
In pooled analyses for all countries, we found a ‘dose–response’
relationship between single motherhood duration and health
(figure 1). For ADL limitations, being a single mother for 1–
3 years was associated with a RR of 1.01 (0.87 to 1.18); while
being a single mother for 14+ years was associated with a RR
of 1.71 (1.49 to 1.97). Divorced single mothers had higher RRs
than widowed single mothers (figure 2A).
Women who were single mothers at younger ages also had
higher RRs for later-life poor health and disability than women
who experienced single motherhood at older ages (figure 2B).
Sensitivity analyses: partnership status, childhood
experiences, health and sociodemographics
A potential concern is the relatively larger proportion of unmarried women with a partner in some European countries.
European women who had a partner during spells of single
motherhood had better health on average than other single
mothers, but worse health than married mothers. However,
effect estimates were imprecise and CIs (etable 1). Sensitivity
analyses indicate that risks between lone motherhood and outcomes are very similar regardless of whether we define lone
motherhood by marital status alone, or include non-marital
partnership in the definition. This is true for Scandinavia and
other regions. For Scandinavian countries in SHARE, ADL risks
associated with single motherhood were 1.56 (CI 1.11 to 2.18)
when based on marital status alone, compared to 1.50 (CI 1.10
to 2.05) when defined by marital status and non-marital partnership. Risk ratios for IADL were 0.99 (CI 0.62 to 1.58) for
Berkman LF, et al. J Epidemiol Community Health 2015;0:1–8. doi:10.1136/jech-2014-205149
3
Research report
Table 2 Demographic and health characteristics by single motherhood experience among mothers aged 50+ by region
Mean age (years)
Married mother
Ever single mother
No (%) Currently married
Married mother
Ever single mother
Number of children=2
Married mother
Ever single mother
Number of children ≥3
Married mother
Ever single mother
Secondary education
Married mother
Ever single mother
Primary education or less
Married mother
Ever single mother
Bottom income quintile
Married mother
Ever single mother
Top income quintile
Married mother
Ever single mother
Bottom wealth quintile
Married
Ever single mother
Top wealth quintile
Married mother
Ever single mother
No (%) Any ADL limitations
Married mother
Ever single mother
Any IADL limitations
Married mother
Ever single mother
Fair/poor SRH
Married mother
Ever single mother
USA
(n=17 866)
England
(n=6294)
Scandinavia
(n=2972)
Western Europe
(n=8576)
Southern Europe
(n=5305)
Eastern Europe
(n=1817)
67.1
62.5
66.5
62.4
67.7
61.8
65.3
63.7
66.1
65.6
64.5
62.7
7569 (67.4)
2078 (39.3)
3226 (77.2)
305 (26.7)
1445 (65.7)
594 (45.5)
5086 (69.1)
636 (32.8)
3732 (70.1)
123 (25.0)
1080 (63.2)
107 (28.0)
3978 (37.4)
1502 (31.3)
2350 (47.5)
496 (35.2)
950 (49.0)
458 (41.5)
2821 (41.3)
624 (32.6)
2276 (42.1)
160 (27.5)
733 (41.9)
120 (36.6)
6291 (50.4)
3069 (51.8)
1601 (32.9)
543 (43.5)
676 (36.1)
364 (33.4)
2744 (38.6)
665 (40.7)
1794 (43.0)
174 (45.4)
555 (45.8)
123 (42.9)
4741 (40.2)
1958 (36.2)
1789 (33.7)
504 (33.6)
553 (28.1)
347 (32.3)
2249 (39.0)
575 (42.4)
763 (12.8)
86 (16.3)
610 (41.8)
126 (45.4)
2421 (17.4)
1522 (22.5)
2632 (57.9)
719 (59.6)
774 (43.6)
364 (36.1)
3277 (42.0)
761 (37.8)
3686 (82.8)
346 (77.8)
789 (52.3)
164 (51.4)
2109 (15.7)
1820 (30.2)
742 (16.1)
328 (26.9)
280 (18.4)
149 (15.6)
1178 (17.2)
388 (23.8)
889 (18.9)
116 (23.8)
257 (19.3)
85 (27.2)
2016 (21.3)
657 (14.9)
1133 (22.0)
232 (15.5)
469 (21.4)
264 (22.8)
1504 (21.0)
292 (17.3)
1011 (19.5)
94 (20.0)
342 (22.2)
46 (12.2)
1787 (13.7)
1955 (33.8)
611 (14.1)
454 (37.9)
210 (12.8)
209 (22.1)
946 (14.2)
469 (30.4)
810 (17.6)
130 (31.9)
226 (18.3)
65 (22.8)
2543 (22.7)
549 (11.7)
1187 (22.4)
172 (10.9)
509 (25.0)
234 (20.6)
1592 (23.5)
250 (15.1)
980 (19.5)
71 (16.1)
322 (21.9)
55 (16.1)
1863 (15.0)
1089 (18.9)
874 (18.4)
329 (26.3)
142 (8.4)
90 (9.5)
590 (9.7)
154 (10.5)
495 (12.3)
64 (14.3)
233 (21.2)
39 (17.8)
1711 (13.8)
929 (16.1)
539 (11.9)
236 (20.0)
119 (7.5)
50 (5.5)
445 (7.6)
107 (6.7)
402 (9.9)
50 (11.1)
178 (16.6)
28 (11.6)
3098 (23.9)
2020 (34.4)
1272 (27.2)
524 (42.9)
384 (21.2)
240 (23.0)
1941 (33.1)
580 (40.7)
2060 (49.5)
207 (52.3)
769 (59.6)
166 (57.6)
Mothers include women who have ever had any biological or adopted children. A woman was defined as having been ‘ever single mother’ if, in any year when she was age 15–49, she
had a child when under the age of 18 years but was not married. ‘Married mothers’ were continuously married in all years they had children when under the age of 18 years, that is,
they were never single mothers.
‘(I)ADLs’: (Instrumental) activities of daily living.
Number of observations and the percentages (in parentheses) are shown. Sampling weights are used for estimating percentages.
Data sources: HRS in year 2004 and 2006; ELSA in year 2004 and 2006; SHARE in year 2004 and 2006, and SHARELIFE in year 2008.
Countries in each regions are: Scandinavia (Denmark and Sweden), Western Europe (Austria, France, Germany, Switzerland, Belgium and the Netherlands), Southern Europe (Italy, Spain
and Greece) and Eastern Europe (Poland and the Czech Republic).
ADLs, activities of daily living; IADLs, instrumental ADLs; ELAS, English Longitudinal Study of Aging; HRS, Health and Retirement Study; SHARE, Survey of Health, Aging, and Retirement
in Europe
lone mothers (including partners) compared to 0.98 (CI 0.63 to
1.53) for lone mothers without partners; and risk ratios for
poor SRH were 1.15(CI 0.95 to 1.40) for unmarried mothers
compared to 1.20 (CI 1.01 to 1.44) for unmarried mothers
without partners. Numbers are smaller in these analyses and CIs
wider than in pooled analyses, yet our ‘bottom line’ is that partnership status does not substantively change our findings. HRS
did not assess partnership.
Given the concern about selection into single motherhood, we
assess whether social or health circumstances in childhood
4
influence risk of becoming a single mother. Among European
women, single mothers averaged worse childhood health and SES
(etable 2). Adjusting for these factors did not substantially change
associations between single motherhood and health outcomes in
England or Western Europe, but did attenuate associations with
ADLs in Scandinavia (etable 3). Results suggest that some—
although not all—cross-national variations may reflect differences
in mechanisms leading to selection into single motherhood.
Women with single-motherhood histories had greater cardiovascular risks than others (etable 4). Adjustment for potential
Berkman LF, et al. J Epidemiol Community Health 2015;0:1–8. doi:10.1136/jech-2014-205149
Research report
Table 3 Single motherhood and adjusted relative risks of disability and SRH among mothers aged 50+ by region
Model specification I
Any ADLs limitations
RR (95% CI)
(1)
Model specification II
Any IADLs limitations
RR (95% CI)
(2)
Fair/poor SRH
RR (95% CI)
(3)
Ever single mother in
USA
1.27***
1.27***
1.32***
(1.14 to 1.40)
(1.14 to 1.42)
(1.22 to 1.42)
England
1.51***
1.66***
1.61***
(1.29 to 1.77)
(1.36 to 2.02)
(1.43 to 1.81)
Scandinavia
1.50*
0.98
1.20*
(1.10 to 2.05)
(0.63 to 1.53)
(1.01 to 1.44)
Western Europe
1.09
0.96
1.23***
(0.80 to 1.47)
(0.66 to 1.39)
(1.09 to 1.39)
Southern Europe
1.13
1.07
1.06
(0.80 to 1.60)
(0.73 to 1.55)
(0.92 to 1.22)
Eastern Europe
0.93
0.87
1.01
(0.66 to 1.31)
(0.59 to 1.28)
(0.89 to 1.15)
p Value for Wald test of equivalence of relative risks associated with single motherhood
All six regions
0.074
0.011
0.000
England vs USA
0.068
0.021
0.004
England vs Scandinavia
0.971
0.036
0.008
England vs Western Europe
0.061
0.011
0.002
England vs Southern Europe
0.139
0.042
0.000
England vs Eastern Europe
0.012
0.003
0.000
Any ADLs limitations
RR (95% CI)
(4)
Any IADLs limitations
RR (95% CI)
(5)
Fair/poor SRH
RR (95% CI)
(6)
1.09
(0.99 to 1.21)
1.40***
(1.19 to 1.65)
1.40*
(1.03 to 1.90)
0.99
(0.73 to 1.35)
0.97
(0.69 to 1.38)
0.90
(0.64 to 1.27)
1.09
(0.98 to 1.21)
1.41***
(1.15 to 1.71)
0.87
(0.57 to 1.33)
0.87
(0.60 to 1.26)
0.94
(0.65 to 1.37)
0.84
(0.58 to 1.22)
1.16***
(1.09 to 1.25)
1.47***
(1.31 to 1.66)
1.12
(0.93 to 1.34)
1.17*
(1.04 to 1.32)
1.04
(0.91 to 1.20)
1.01
(0.89 to 1.15)
0.037
0.011
0.991
0.051
0.064
0.024
0.046
0.027
0.046
0.025
0.063
0.016
0.000
0.001
0.012
0.008
0.000
0.000
***p<0.001, **p<0.01, *p<0.05.
Each column from Column (1) to (6) presents adjusted relative risks (point estimates and 95% CIs) of single motherhood by region, obtained from modified Poisson regressions, with
robust variance clustered at the individual level. Data are weighted by sampling weights. Other control variables in ‘Model specification I’ include the interactions of six country/region
dummies with age, age squared, secondary education, primary education or less, number of children and current marital status, as well as country- and time-fixed effects. Extra control
variables are in ‘Model specification II’: per-capita household income quintiles and per-capita household wealth quintiles. Wald test p values correspond to the null hypothesis that the
relative risks associated with single motherhood are the same in the regions specified.
Data sources: HRS in year 2004 and 2006; ELSA in year 2004 and 2006; SHARE in year 2004 and 2006, and SHARELIFE in year 2008.
Countries in each regions are: Scandinavia (Denmark and Sweden), Western Europe (Austria, France, Germany, Switzerland, Belgium and the Netherlands), Southern Europe (Italy, Spain
and Greece) and Eastern Europe (Poland and the Czech Republic).
ADLs, activities of daily living; IADLs, instrumental ADLs; ELAS, English Longitudinal Study of Aging; HRS, Health and Retirement Study; SHARE, Survey of Health, Aging, and Retirement
in Europe; SRH, self-rated health.
Figure 1 Single motherhood and
adjusted relative risks of disability and
SRH among mothers aged 50+, by
quartiles of single motherhood
duration. ‘*’ Indicates p value <0.05.
Data from ELSA, SHARE and HRS are
pooled in this analysis. Adjusted
relative risks were obtained from
modified Poisson regressions, with
robust variance clustered at the
individual level. Key independent
variables include binary indicators on
quartiles of single motherhood
duration: 1–3, 4–7, 8–13, 14+ years.
The comparison group is mothers who
never had single motherhood
experience before age 50. Additional
covariates include age, age squared,
education, number of children, current
marital status, time of interview and
country-fixed effects. Data are
weighted by sampling weights. ELAS,
English Longitudinal Study of Aging;
HRS, Health and Retirement Study;
SHARE, Survey of Health, Aging, and
Retirement in Europe; SRH, self-rated
health.
Berkman LF, et al. J Epidemiol Community Health 2015;0:1–8. doi:10.1136/jech-2014-205149
5
Research report
Figure 2 Single motherhood and
adjusted RRs of disability and SRH
among mothers aged 50+, by causes
or ages of single motherhood,
conditional on single motherhood
duration. ‘*’ Indicates p value <0.05.
Adjusted RRs of single motherhood by
causes and ages of the single
motherhood experience were estimated
by running two sets of Poisson
regressions: one has causes and
quartiles of single motherhood
duration as key independent variables,
while the other has ages and quartiles
of single motherhood duration as key
independent variables. RRs reported in
(A and B) reflect the RRs of single
mothers assuming a single
motherhood duration of 8–13 years.
The comparison group is mothers who
never had single motherhood
experience before age 50. Additional
covariates include age, age squared,
education, number of children, current
marital status, time of interview and
country-fixed effects. ADLs, activities
of daily living; IADLs, instrumental
ADLs; SRH, self-rated health; RH,
relative risks.
mediators, including smoking (etable 5), obesity (etable 5) and
hypertension, plus diabetes, stroke or heart disease (etable 6),
attenuated but did not eliminate associations between single
motherhood and poor health. Additional sensitivity analyses are
shown in etables 7–12.
DISCUSSION
Lifetime experiences of single motherhood were associated with
increased risks of physical limitations and poor health at older
ages among mothers in England, Scandinavian countries and the
USA. Single motherhood was less consistently associated with
health in Continental Western, Eastern and Southern European
countries. Longer duration of single motherhood was associated
with poorer outcomes.
Potential explanations for association between single
motherhood and later-life health
Controlling for income and wealth attenuated effects in the
USA, but less so in other regions. Associations may reflect
6
selection and causation in cycles of disadvantage: poverty
increasing risk of single motherhood reflecting, in part, earlier
health disadvantages. Being a lone mother may hamper
women’s abilities to gain education, accrue careers and
accumulate income, also leading to poorer health. While our
study is longitudinal in design, we often draw on retrospective
recall of events occurring in early adulthood. Longitudinal
data, prospectively following women from early to late adulthood, would better enable us to disentangle pathways and
mechanisms.
Single motherhood was strongly associated with adverse
health in Sweden and Denmark. Two previous studies have
shown that current single mothers in Sweden as well as in
Britain had higher prevalence of poor SRH and chronic illnesses relative to coupled mothers, and magnitudes of relative
differences were similar for these countries.4 9 In our study,
although adjusting for later-life socioeconomic conditions
somewhat attenuated RRs for Scandinavia (7%), associations
between single motherhood and ADL limitations remained
Berkman LF, et al. J Epidemiol Community Health 2015;0:1–8. doi:10.1136/jech-2014-205149
Research report
statistically significant. Strikingly, associations between single
motherhood, and ADLs and SRH in the USA and Scandinavia,
were similar. We do present multiplicative effect estimates
(RRs), so this result should be interpreted in light of overall
better health in Scandinavia. Nevertheless, mechanisms besides
poor social protection policies, such as a lack of social or
family support, may have contributed to this finding. Future
studies should incorporate employment experiences since it is
likely that employment contributes to long-run health and may
relate to single motherhood. Detailed work histories necessary
for these analyses are not available in a comparable way across
all countries included here, prohibiting a full analysis. We
acknowledge that employment patterns may be an explanation
for observations.
Across all regions, single mothers were more likely to be
smokers but not more likely to be obese; controlling for
these risks did not eliminate associations. Controlling for
cardiovascular conditions moderately reduced RRs in the
USA/England, but not in other European countries, suggesting
that such conditions may partially explain links between
single motherhood and later functional impairment.
The role of social support in shaping observed associations
Social support and cohesive networks may partially explain
associations between single motherhood and health. Social
support is itself an important predictor of adult health and
functioning.36–39 Although we did not have detailed data in
mid-life, social support might play an important role in alleviating strains of single motherhood. For example, in Southern
Europe, a region emphasising family solidarity, single motherhood is not associated with increased health risks. In the USA,
where Hispanics tend to have more family support than
non-Hispanic whites,40 Hispanic single mothers did not have
increased risks.
Our results identify several vulnerable populations. Women
with prolonged spells of single motherhood; those whose single
motherhood resulted from divorce; women who became single
mothers at young ages; and single mothers with two or more
children, were at particular risk.
Strengths and limitations
Major strengths are harmonised data across many countries,
and in-depth retrospective data on marriage and childbearing.
The greatest limitation is reliance on self-reported health
outcomes. Although SRH is a general health measure, it has
been repeatedly shown to predict objective outcomes such as
mortality.41 ADLs and IADLs are commonly assessed by
self-report.
HRS did not collect retrospective data on non-marital or
same-sex partnership, so we were not able to test whether these
partnerships offered similar protections as marriage. In sensitivity analyses, we found that incorporating non-married partners
into analyses for Scandinavian and other European countries
did not change associations substantially. Finally, we did not
have retrospective information on SES, social support or networks during single motherhood, so we cannot explicitly
examine roles of these conditions during childbearing years in
shaping observed effects.
The risks observed in Scandinavian countries are provocative
and we speculate about some reasons for increased risks.
However, previous research focusing on health inequalities have
also found that Scandinavia (Denmark, Finland, Norway,
perhaps with the exception of Sweden) have larger inequalities
in mortality by educational attainment than other European
countries, and particularly in Southern European, where
inequalities tend to be smaller despite less generous welfare state
traditions. Thus, while surprising, our study is not contradictory
to previous evidence that countries with generous welfare states
may have smaller income inequality, but not necessarily smaller
health inequality.42 A second issue relates to the risk of poverty
among single mothers. In general, given higher levels of income
support policies and overall lower levels of poverty in
Scandinavia compared to other countries, it seems unlikely that
poverty would be more strongly associated with lone motherhood in Scandinavian countries compared to other countries.
A third explanation refers to the role of employment. Indeed,
employment rates in Sweden and Denmark were relatively high
compared to rates in other countries. It is difficult to predict
whether this would result in larger or smaller risks associated
with single motherhood. For example, higher employment
rates among lone mothers may reduce poverty rates for
Scandinavian women, leading to smaller health risks associated
with lone motherhood. On the other hand, higher levels of
stress in combining work and family roles may have increased
work-family strain, potentially leading to worse health. In addition, issues raised with regard to feminist theories about the
welfare state may be important.16 17 For example, Sweden had
more generous maternity benefits than other countries during
the time that many women were single mothers in this study. It
is possible that these policies reinforced the gendered division
of roles and the strain associated with continued unpaid caregiving coupled with labour force participation. Although speculative, higher rates of work-family conflict may have contributed
to their higher risk of poor health in later life. Finally, studies of
social isolation suggest that risks of social isolation may be
greater in Sweden—similar to US rates. Family dynamics and
informal support in Southern Europe may play a protective
role.
Conclusions and future directions
Findings add to the growing recognition that single motherhood may have long-term health effects on mothers.6 10 43 As
lone motherhood is on the rise in many countries, policies
addressing health disadvantages of lone mothers may be
essential to improving women’s health and reducing disparities. Social support and family dynamics may further protect
single mothers. In environments where social interactions are
valued at a cultural level, we find reduced risks. Anti-poverty
programmes may additionally moderate impacts of single parenting. Access to family planning resources and policies that
help single mothers remain in the labour force, balancing
work and family demands, as well as informal work-family
practices, may yield important benefits for single mothers and
their families.
What is already known on this subject?
Single motherhood has been shown to predict multiple
health conditions, concurrently with single motherhood
and also later in life. While a few studies have compared
two or three countries, no study has explicitly conducted
a cross-national comparison of the magnitude of
the association between single motherhood and later-life
health.
Berkman LF, et al. J Epidemiol Community Health 2015;0:1–8. doi:10.1136/jech-2014-205149
7
Research report
16
17
What does this study add?
18
Lifetime history of single motherhood was associated with
increased risk of later-life disability and poor health in the USA,
the UK and Scandinavia, but not in continental Western, Eastern
or Southern Europe. As prevalence of single motherhood is on
the rise across the developed world, social policies that protect
women in vulnerable family situations may help improve
population health and reduce health disparities as women age.
Contributors YZ conducted the statistical analyses. All authors (LFB, YZ, MMG,
MA, ABS, ELS) conceived the idea of the paper, read, commented and contributed
to the writing of the paper, and approved the final version.
Funding This study was supported by the National Institute on Aging 1 R01
AG040248-02.
Competing interests None declared.
19
20
21
22
23
24
25
Ethics approval University of Michigan (HRS); London Multi-centre Research Ethics
Committee (ELSA); University of Mannheim (SHARE).
26
Provenance and peer review Not commissioned; externally peer reviewed.
27
Data sharing statement All data from ELSA, SHARE and HRS are publicly
available from the relevant cohorts.
28
REFERENCES
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
8
Benzeval M. The self-reported health status of lone parents. Soc Sci Med
1998;46:1337–53.
Berkman PL. Spouseless motherhood, psychological stress, and physical morbidity. J
Health Soc Behav 1969;10:323–34.
Burstrom B, Diderichsen F, Shouls S, et al. Lone mothers in Sweden: trends in
health and socioeconomic circumstances, 1979–1995. J Epidemiol Community
Health 1999;53:750–6.
Burstrom B, Whitehead M, Clayton S, et al. Health inequalities between lone and
couple mothers and policy under different welfare regimes—the example of Italy,
Sweden and Britain. Soc Sci Med 2010;70:912–20.
Fritzell S, Burstrom B. Economic strain and self-rated health among lone and couple
mothers in Sweden during the 1990s compared to the 1980s. Health Policy
2006;79:253–64.
Grundy EMD, Tomassini C. Marital history, health and mortality among older men
and women in England and Wales. BMC Public Health 2010;10.
Ringback Weitoft G, Haglund B, Rosen M. Mortality among lone mothers in
Sweden: a population study. Lancet 2000;355:1215–19.
Weitoft GR, Haglund B, Hjern A, et al. Mortality, severe morbidity and injury among
long-term lone mothers in Sweden. Int J Epidemiol 2002;31:573–80.
Whitehead M, Burstrom B, Diderichsen F. Social policies and the pathways to
inequalities in health: a comparative analysis of lone mothers in Britain and
Sweden. Soc Sci Med 2000;50:255–70.
Williams K, Sassler S, Frech A, et al. Nonmarital Childbearing, Union History, and
Women’s Health at Midlife. Am Sociol Rev 2011;76:465–86.
Young LE, Cunningham SL, Buist DSM. Lone mothers are at higher risk for
cardiovascular disease compared with partnered mothers. Data from the National
Health and Nutrition Examination Survey III (NHANES III). Health Care Women Int
2005;26:604–21.
Montez J, Sabbath E, Glymour M, et al. Trends in work-family context among U.S.
women by education level, 1976 to 2011. Popul Res Policy Rev 2014;33(5):629–48.
Gornick J, Meyers M. More alike than different: revisiting the long- term prospects
for developing “European- Style” work/family policies in the United States. J Com
Policy Anal Res Pract 2004;6:251–73.
Gornick JC, Meyers M. Families that work: policies for reconciling parenthood and
employment. New York: Russell Sage Foundation, 2003.
Ruspini E. Living on the poverty line: lone mothers in Belgium, Germany, Great
Britain, Italy and Sweden. MZES Working Papers, 1998.
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
Orloff A. Gender in the Welfare State. Annu Rev Sociol 1996;22:51–78.
Lewis J. Gender and Welfare Regimes: Further Thoughts. Soc Polit
1997:160–77.
Börsch-Supan A, Jürges H. The survey of health, ageing and retirement in Europe—
methodology. Mannheim: MEA, 2005.
Scholes S, Medina J, Cheshire H, et al. Technical report (wave 3): living in the 21st
century: older people in England: the 2006 English Longitudinal Study of Ageing.
National Centre for Social Research, 2009.
Schröder M. Retrospective data collection in the survey of health, ageing and
retirement in Europe. SHARELIFE Methodology. Mannheim: MEA, 2011.
Servais M. Overview of HRS Public Data Files for Cross-sectional and Longitudinal
Analysis. 2010.
Meijer E, Zamarro G, Fernandes M. Comparison between SHARE, ELSA and HRS.
In: Börsch-Supan A, Brugiavini A, Jürges H, et al., eds. Health, Ageing and
Retirement in Europe (2004–2007)—Starting the longitudinal dimension.
Mannheim: MEA, 2008:99–107.
Health and Retirement Study. Sample sizes and response rates. University of
Michigan, 2011.
Katz S, Ford AB, Moskowitz RW, et al. Studies of illness in the aged: the index of
ADL: a standardized measure of biological and psychosocial function. JAMA
1963;185:914–19.
Guralnik JM, LaCroix AZ, Abbott RD, et al. Maintaining Mobility in Late
Life. I. Demographic Characteristics and Chronic Conditions. Am J Epidemiol
1993;137:845–57.
Zhang J, Yu KF. What’s the relative risk? A method of correcting the odds ratio in
cohort studies of common outcomes. JAMA 1998;280:1690–1.
Brown TT, Cole SR, Li X, et al. Antiretroviral therapy and the prevalence and
incidence of diabetes mellitus in the multicenter AIDS cohort study. Arch Intern Med
2005;165:1179–84.
LaCoursiere DY, Bloebaum L, Duncan JD, et al. Population-based trends and
correlates of maternal overweight and obesity, Utah 1991–2001. Am J Obstet
Gynecol 2005;192:832–9.
Pollack AZ, Buck Louis GM, Sundaram R, et al. Caffeine consumption and
miscarriage: a prospective cohort study. Fertil Steril 2010;93:304–6.
Szklo-Coxe M, Young T, Peppard PE, et al. Prospective associations of insomnia
markers and symptoms with depression. Am J Epidemiol 2010;171:709–20.
Zou GY. A modified Poisson regression approach to prospective studies with binary
data. Am J Epidemiol 2004;159:702–6.
Yelland LN, Salter AB, Ryan P. Performance of the Modified Poisson Regression
Approach for Estimating Relative Risks From Clustered Prospective Data. Am J
Epidemiol 2011;174:984–92.
Spiegelman D, Hertzmark E. Easy SAS Calculations for Risk or Prevalence Ratios and
Differences. Am J Epidemiol 2005;162:199–200.
Judge GG. The Theory and practice of econometrics. New York: Wiley, 1985.
The Organisation for Economic Co-operation and Development (OECD). Growing
Unequal? Income Distribution and Poverty in OECD Countries. Paris: OECD, 2008.
Berkman LF, Melchior M, Chastang JF, et al. Social integration and mortality:
a prospective study of French employees of Electricity of France-Gas of France: the
GAZEL Cohort. Am J Epidemiol 2004;159:167–74.
Berkman L, Glass TA. Social integration, social networks, social support, and health.
In: Berkman L, Kawachi I, eds. Social Epidemiology. New York: Oxford University
Press, 2000:137–73.
Capistrant BD, Moon JR, Berkman LF, et al. Current and long-term spousal
caregiving and onset of cardiovascular disease. J Epidemiol Community Health
2012;66:951–6.
Kumar S, Calvo R, Avendano M, et al. Social support, volunteering and health
around the world: cross-national evidence from 139 countries. Soc Sci Med
2012;74:696–706.
Almeida J, Molnar BE, Kawachi I, et al. Ethnicity and nativity status as determinants
of perceived social support: testing the concept of familism. Soc Sci Med
2009;68:1852–8.
Idler EL, Benyamini Y. Self-rated health and mortality: a review of twenty-seven
community studies. J Health Soc Behav 1997;38:21–37.
Mackenbach JP. The persistence of health inequalities in modern welfare states:
the explanation of a paradox. Soc Sci Med 2012;75:761–9.
Williams K, Sassler S, Nicholson LM. For better or for worse? The consequences
of marriage and cohabitation for single mothers. Soc Forces
2008;86:1481–511.
Berkman LF, et al. J Epidemiol Community Health 2015;0:1–8. doi:10.1136/jech-2014-205149